Scholarlink

As global supply chains experience rapid digital transformation, the ability to forecast demand accurately has become a strategic necessity. This paper provides a structured review of 37 peer-reviewed studies published between 2020 and 2025 that investigate the use of machine learning (ML) techniques in demand forecasting within supply chains. The findings show that while deep learning models continue to be widely employed for demand forecasting, models such as Random Forest, Support Vector Machines (SVM), and XGBoost are increasingly gaining traction due to their robustness, computational efficiency, and interpretability. Additionally, the study identifies critical shortcomings, including a lack of explainable AI (XAI) integration, inconsistencies in benchmarking, and limited evaluation across diverse datasets. Future research directions are suggested, focusing on integrating XAI techniques, establishing standardized hybrid model frameworks, and exploring Transformer architectures for real-time, multi-signal forecasting. These insights provide valuable guidance for researchers and practitioners seeking to implement intelligent, context-aware forecasting systems in digitally driven trade environments.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Learning the Chain: A Review of Predictive Machine Learning Models in Modern Supply Chain

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Ahmad N. Shawaqfeh,
  • Juan F. De Paz

摘要

As global supply chains experience rapid digital transformation, the ability to forecast demand accurately has become a strategic necessity. This paper provides a structured review of 37 peer-reviewed studies published between 2020 and 2025 that investigate the use of machine learning (ML) techniques in demand forecasting within supply chains. The findings show that while deep learning models continue to be widely employed for demand forecasting, models such as Random Forest, Support Vector Machines (SVM), and XGBoost are increasingly gaining traction due to their robustness, computational efficiency, and interpretability. Additionally, the study identifies critical shortcomings, including a lack of explainable AI (XAI) integration, inconsistencies in benchmarking, and limited evaluation across diverse datasets. Future research directions are suggested, focusing on integrating XAI techniques, establishing standardized hybrid model frameworks, and exploring Transformer architectures for real-time, multi-signal forecasting. These insights provide valuable guidance for researchers and practitioners seeking to implement intelligent, context-aware forecasting systems in digitally driven trade environments.